ONNX weight is a kind of deep learning model representation format, ONNX format allows the AI developers in the transformation between different framework model, to call on the generality and the current PyTorch *, Caffe2 *, Apache MXNet *, Microsoft Cognitive Toolkit *, baidu fly blade support ONNX format, OpenVINO model optimizer support the ONNX intermediary file format of the model transformation IR,
Current OpenVINO official support ONNX model mainly includes: bert_large, bvlc_alexnet, bvlc_googlenet, bvlc_reference_caffenet, bvlc_reference_rcnn_ilsvrc13 model, inception_v1, inception_v2, resnet50, squeezenet, densenet121, emotion_ferplus, mnist, shufflenet, VGG19, zfnet512, it is important to note these model upgrade version is not supported,
Since 2019 r04 OpenVINO version supports all Pytorch model of public support for the model list as follows:
Pytorch ONNX conversion to OpenVINO IR
This example demonstrates how to convert ONNX from torchvision public model, and then converted to IR, use OpenVINO with the complete process of call, we will with resnet18 as an example to demonstrate,
download model and turn ONNX format
To download and use torchvision training model, first need to install good pytorch, and then execute the following code can download related support model:
1 import torchvision. Models as models
2 resnet18=models. Resnet18 (pretrained=True)
3 alexnet=models. Alexnet (pretrained=True)
4 squeezenet=models. Squeezenet1_0 (pretrained=True)
5 vgg16=models. Vgg16 (pretrained=True)
6 densenet=models. Densenet161 (pretrained=True)
Inception=7 models. Inception_v3 (pretrained=True)
8 googlenet=models. Googlenet (pretrained=True)
9 shufflenet=models. Shufflenet_v2_x1_0 (pretrained=True)
Mobilenet=10 models. Mobilenet_v2 (pretrained=True)
11 resnext50_32x4d=models. Resnext50_32x4d (pretrained=True)
12 wide_resnet50_2=models. Wide_resnet50_2 (pretrained=True)
13 mnasnet=models. Mnasnet1_0 (pretrained=True)
Here, we only need to perform resnet18=models. Resnet18 (pretrained=True) resnet18 can be downloaded on the model, the model of input format requirements as follows:
Size is 224 x224,
RGB three-channel images,
Mean=[0.485, 0.456, 0.406]
STD=[0.229, 0.224, 0.225]
Download and turned as ONNX code is as follows:
model=torchvision. Models. Resnet18 (pretrained=True). The eval ()
Dummy_input=torch. Randn ((1, 3, 224, 224))
The torch. Onnx. Export (model, dummy_input, "resnet18. Onnx")
to the IR format
Cmd to open the installed OpenVINO:
Deployment_tools \ model_optimizer
Directory, execute the following command line statement:
python mo_onnx. Py -- -- input_model D: \ \ python pytorch_tutorial \ resnet18 onnx
You can see resnet18 model has better success!
OpenVINO SDK call
To convert good IR model, you can first by OpenVINO202R3 Python version of the SDK to complete accelerating reasoning forecast, the complete code implementation is as follows:
the from __future__ import print_function
The import cv2
The import numpy as np
The import of logging as the log
The from openvino. Inference_engine import IECore
With the open (' imagenet_classes. TXT) as f:
Labels=[line strip () for the line in f.r eadlines ()]
Def image_classification () :
Model_xml="resnet18. XML"
Model_bin="resnet18. Bin"
# Plugin initialization for specified device and the load extensions library if specified
The info (" Creating Inference Engine ")
Ie=IECore ()
# Read IR
The info (" Loading network files: \ n \ t \ n \ t {} {} ". The format (model_xml model_bin))
Net=ie. Read_network (model=model_xml, weights=model_bin)
The info (" Preparing input blobs ")
Input_blob=next (iter (net. Inputs))
Out_blob=next (iter (net. Outputs))
# Read and pre - process input images'
N, c, h, w=net. Inputs [input_blob]. Shape
Images=np. Ndarray (shape=(n, c, h, w))
SRC=https://bbs.csdn.net/topics/cv2.imread (" D:/images/messi. JPG ")
Image=cv2. Resize (SRC (w, h))
Image=np. Float32 (image)/255.0
Image [: :] -=(np) float32 (0.485), np. Float32 (0.456), np. Float32 (0.406))
Image [: :]/=(np) float32 (0.229), np. Float32 (0.224), np. Float32 (0.225))
Image=image transpose ((2, 0, 1))
# Loading model to the plugin
The info (" Loading model to the plugin ")
Exec_net=ie. Load_network (network=net, device_name="CPU")
# Start sync inference
The info (" Starting inference in synchronous mode ")
Res=exec_net. Infer (inputs={input_blob: [image]})
# Processing output blob
The info (" Processing output blob ")
Res=res [out_blob]
Label_index=np. Argmax (res, 1)
Label_txt=labels [label_index [0]]
Cv2. PutText (SRC, label_txt, (10, 50), cv2. FONT_HERSHEY_SIMPLEX, 1.0, (255, 0, 255), 2, 8)
Cv2. Imshow (" ResNet18 -from Pytorch image classification ", SRC)
Cv2. WaitKey (0)
Cv2. DestroyAllWindows ()
If __name__=="__main__ ':
Image_classification ()
The results are as follows:
